Fiber optics communication systems require compact light emitting sources capable of generating single-mode, tunable, narrow linewidth radiation in the 1.3-1.56 .mu.m wavelength range. Some of the existing semiconductor lasers, for example, InGaAsP DFB lasers can meet requirements for high power and proper wavelength, but fail to satisfy requirements for high side mode suppression ratio (SMSR), predictability and controllability of generated wavelength, insensitivity to external feedback and random facet phase variations, simple manufacturing and high device yield. Additionally, a rapid advance in high speed and large capacity dense wavelength division multiplexing (DWDM) fiber optics systems continues to demand semiconductor lasers not only possessing properties mentioned above but also capable of providing a wide continuous tuning range and multi-wavelength generation for practical and cost effective applications.
Conventional index coupled DFB lasers employing an index corrugation have an inherent problem in existence of two longitudinal modes with an equal threshold gain which results in poor single mode operation as shown, for example, in the article by H. Kogelnik and C. V. Shank "Coupled-mode theory of distributed feedback lasers", J. Appl. Phys., vol. 43, no. 5, pp. 2327-2335, 1972.
For index coupled DFB lasers, the longer and shorter wavelength Bragg modes around the laser stop band are intrinsically degenerate in terms of the threshold gain. The degeneracy may be broken, for example, in the presence of asymmetric facet coatings and facet phase variations. The yield of DFB lasers with a fixed lasing wavelength and a predetermined SMSR is very low in practice because of the random variations of facet phase, yield usually being not more than several percent. Without internal built-in mode discrimination between the two degenerate modes, mode properties of index coupled DFB lasers are primarily determined by asymmetric facet coatings and variations of facet phases. As a result these lasers are very sensitive to variations of the effective laser facet phases and can be strongly influenced by any external feedback.
For quarter wavelength shifted DFB lasers, described, for example, in the article by K. Utaka, S. Akiba, K. Sakai, and Y. Matsushima ".lambda./4-shifted InGaAsP/InP DFB lasers",IEEE J. of Quantum Electronics, Vol. QE-22, no.7, p.1042-1052 (1986), an additional phase shift is introduced into the laser structure to break the degeneracy between the two Bragg modes around the stop band. The yield of this type of lasers ensuring single-mode operation is higher than the conventional index coupled DFB lasers. However, since the laser operation is based upon an additional phase introduced into the structure, it is critically dependent on the phase shift which has been actually introduced to the laser and which is very difficult to control or manufacture in large scale. The laser facet phase still plays a significant role since it works together with the introduced built-in phase to satisfy the round-trip phase condition for resonance. A good anti-reflection (AR/AR) coating can reduce the effect of the facet phase on the laser properties. However, the quarter wavelength shift laser usually suffers from a large longitudinal spatial-hole burning (SHB), resulting from the phase shift introduced in the center of the laser, especially when a large index coupling is required to reduce a threshold current in the case of AR/AR coating. Strong SHB may quickly degrade the SMSR when the injection current is increased.
When a laser is facet phase sensitive or critically dependent on the phase shift introduced within the cavity, it becomes very sensitive to any perturbations or variations in its operation conditions. When a number of such lasers are arranged in a series, they interact with each other. The presence of one laser influences the operation of other lasers. One laser usually acts as an effective grating-based reflector to cause reflections fed back into other lasers to vary not only in amplitude, but also in phase, both being wavelength dependent. Additionally, both the amplitude and the phase are also dependent on the operation conditions of adjacent lasers, such as temperature, injection current and leakage current between the lasers. Thus, interaction between lasers significantly influences lasing behaviour of each laser, resulting in an extremely low device yield and poor laser performance. Often stable operation of the series as a whole is impossible.
O. Sahlen, L. Lundqvist, J. Terlecki and J. P. Weber in the article "A robust WDM network laser source: the DFB-series laser", ThB1, OFC'97, Dallas, USA, described an attempt to use quarter wavelength shifted DFB lasers as building blocks in a series. Although quarter wavelength shifted DFB lasers exhibit a high single mode yield in theory, they suffer from a large spatial hole burning as discussed above. Their current tuning range is also relatively small due to the potential onset of other longitudinal modes caused by spatial hole burning. As a result, the series as a whole did not demonstrate high performance and stability, and simultaneous multi-wavelength operation was not reported at all.
The predictability of the lasing wavelength for each individual DFB laser is also a critical parameter for the series operation. Even if one of the lasers, working as a perfect single-mode laser, happens to lase on a wrong side of the stop band, the entire series will fail in its operation. The same result will happen if one laser unexpectedly switches between the two Bragg modes, which is unacceptable for practical system applications.
Therefore, in order to obtain a good performance of series DFB lasers, it is critical to ensure that each laser operates substantially independently and has no influence on the lasing behaviour of other lasers in the series, thus, providing no substantial interaction between lasers in the series, each laser maintaining high performance characteristics at the same time.